DOCUMENT RESUME

ED 392 256 FL .023 589

AUTHOR Ngonyani, DeoTITLE Towards a Typology of Applicatives in Bantu.PUB DATE Mar 95NOTE 35p.; Paper presented at the Annual Conference on

African Linguistics (26th, Los Angeles, CA, March24-26, 1995).

PUB TYPE Reports Evaluative/Feasibility (142)Speeches/Conference Papers (150)

EDRS PRICE MF01/PCO2 Plus Postage.DESCRIPTORS *African Languages; *Bantu Languages; Classification;

*Language Patterns; Language Research; LinguisticTheory; *Sentence Structure; Swahili; *Syntax;*Uncommonly Taught Languages

IDENTIFIERS Ndendeule

ABSTRACTAn analysis of applicative constructions in Bantu

languages proposes a typology of applicative structures, usingexamples from Ndendeule and Swahili. First, the basic facts aboutapplicative constructions are presented, including those concerningmorphology, meaning, and alternative expressions,'and severalarguments are posited. Primary objedi:s properties diagnostics arethen applied to six different applied objects. Previous typology ofapplied objects is reviewed and the proposed new typology isoutlined. Verb phrase (VP) ellipsis is advanced as evidence of thestructure of the VP, and a complex VP structure for applicativeconstructions is suggested. It is concluded that based on primaryobject properties, there are three types of applicatives whoseprototypes are benefactive, instrumental, and locative. Theseconstituent structures involve two VPs: one VP which contains thetheme/patient is embedded in another VP, which contains the appliedobject. In addition, the applicative morphemE is a predicate headthat selects the minimal VP and the applied object. Implications formapping of the structure are considered briefly. Contains 25refcrences. (MSE)

***********************************************************************

Reproductions supplied by EDRS are the best that can be madefrom the original document.

***********************************************************************

Towards a Typology of Applicatives in Bantu

Deo Ngonyani

PERMISSION TO REPRODUCE THISMATERIAL HAS BEEN GRANTED BY

Qe\-t7.,.fu

TO THE EDUCATIONAL RESOURCESINFORMATION CENTER iERICI

U S DEPARTMENT OF EDUCATIONOlt c, C, F(1,,11.0ra, Reseacr ,,,,,jUCATIONAL RESOURCES INFORMATION

crrCENTER (ERIC)his document has been reproduced aseceived horn the person or organization

onginating it

0 Minor changes have been made to

improve reprocluctIon twallty

Poots ot vIevi or.opintons stated In thsdocvment do not necest,artiv representofficial OERI positnor or policy

26th Annual Conference on Afia,An LinguisticsMarch 24-26, 1995

Los AngelesTowards a Typology of Applicatives in Bantu

Deo Ngonyani

0. IntroductionThe structure of the applicative constructions has been a subject of

intense debate recently. However, it appears there is a lot of confusion aboutthe applied objects. Part of it is due to the fact that applied objects get variousinterpretations and their sentences differ in argument structure as well asother syntactic properties. Generalization about one applied object may notapply to other applied objects. The fact that the objects exhibit differentsyntactic properties suggests different structures are involved. Some appliedobjects, however, seem to show similar syntactic behavior. In this paper I takethe view that in order to understand applicatives in Bantu, it is necessary tolook at their typology. This will enable a person to delimit the relevance ofcertain generalization and to delimit the applicability of some notions onapplicatives. The data for this paper are drawn from Ndendeule1 and Swahili.

This paper is organized into five sections. In Section 1, I present thebasic facts about applicative constructions. I apply the primary objectproperties diagnostics to six different applied objects in Section 2. Section 3discusses previous typology and presents my typology. In Section 4, I use VPellipsis as evidence of the structure of the VP and suggest a complex VPstructure for applicative constructions. I discuss implications of my analysis in

Section 5.

1. Basic FactsBantu languages have an applicative morpheme, {-il-} suffixed to the

verb stem. This suffix increases the number of arguments for the verb by one.Example (1) illustrates this in Ndendeule language2.

1 Ndendeule i closest to Ngindo, P.14 in Guthrie's classification. There is no mention of thislanguage by Guthrie.2 Abbreviations:PRT Present Tense PST Past Tense FUT Future TenseSA Subject Agreement OA Object Agreement FV Final VowelSubj Subject NOM Nominalizer INF InfinitiveBEN Beneficiary INST Instrumental LOC LocativeDO Direct Object JO Indirect Object AO Applied ObjectREC Reciprocal APP Applicative CAUS Causative

1

(1) a. n-gh8ni ngop.o1-guest 1-PST-buy-FV 10cloththe guest bought clothes

b. n-gheni a-ki-n-h8me1-81-a inw-ana ngopo1-guest 1SA-PST-10A-buy-APP-FV 1-child 10cloththe guest bought the child clothes

The contrast between (la) and ( lb) is in the fact that the former has a verbwithout the applicative morpheme and has one object, clothes, while the latterhas the applicative morpheme and two objects, "clothes" and "child". Theadditional object, in this case, "child", is known as the applied object.

1.1 MorphologyThe realization of the applicative morpheme may vary slightly. In

Swahili, it is realized as /-il-/ or depending the vowel height of the preceding

vowel of the verb stem. Consonant final stems will take /-i-/ or /-e-/ againdetermined by vowel harmony. Ndendeule, which has four vowel levels3 hasvowel harmony also and the applicative morpheme is realized as /-il-/, /-el-/ orI-el-I. The vowel systems of Swahili and Ndendeule are as follows:

(2) a. Swahili b. Ndendeule

a a

The applicative morpheme takes the values of the front vowels only. Thuswhen the last vowel of the stem is Iii or /u/, the applicative morpheme will be /-il-/; when it is /el or /o/ in Ndendeule, the applicative will be /-el-/; when it is /e/ or/o/, the applicative will be /-el-/. When the last vowel of the stem is /a/, theapplicative suffix is I-H-1. The following examples illustrate the differentallomorphs of the applicative.

Ndendeule(3) a. -yipa

steal steal from/for

3 For Swahili transcription, I use the standard orthography. But the approximateequivalences of Ndendeule and Swahili vowels are shown in the vowel charts in (2).

2

b. -tula -tulilaskin skin with/for

c. -yembasing

-yembelasing for

d. -pola -polelateach teach for/at

e. -kEma kEmElacall call for

f. -tala -tolclatake take for

g. -kanga -kangilapush push for/to

The following are examples from Swahili.

(4) a. -ticha -fichiahide hide for/at/with

b. -funga -fungiaclose/shut/tie shut for/at/on

c. -peleka -pelekeasend send to/for

d. -omba -ombeaask for something ask for something for

e. -kata katiacut cut for/at/on

Therefore, the applicative morphology in the two languages is the samediffering only in the vowel quality of the suffix.

1.2 MeaningApplicative objects may be assigned different interpretations. The

different interpretations of the applied objects have been recognized bydifferent authors (e.g. Ashton, 1966; Trithart, 1983; Bresnan and Moshi, 1990;Kimenyi, 1980). The following sentences from Ndendeule illustrate sevendifferent interpretations.

3

Ndendeule(5) a. rnayi a-ki-pa-tElek-El-a pa-na ch-akulya (Benefactive)

mother 1SA-PST-20A-cook-APP-FV 2-child 7-foodmother cooked the children some food

b. ma-yanga pa-ki-kayul-il-a ki-peya (Instrumental)6-stone 2SA-PST-break-APP-FV 7-potthe stones, they the pot with them

c. pa-lumba pa-ki-tul-il-a nyama pa-tna-nyabi (Locative)2-hunter 2-PST-skin-APP-FV 9animal 16-6-grassthe hunters skinned the animal on the grass

d. pa-lumba pa-ki-hyem-el-a inbiya (Motive)2-hunter 2SA-PST-hunt-APP-FV 10moneythe hunters hunted for money

e. mbuya a-ki-pa-kang-i f3a-chongo10 pa-chikana (Direction)I grandma 1SA-PST-20A-push-APP 2-boy 2-girlgrandma pushed the girls to the boys

f. ma-yani ya-ki-f3a-yornol-cl-a ma-chi f3a-lumba (Malefactive)6-baboon 6SA-PST-20A-finish-APP-FV 6-water 2-hunterthe baboons fr;shed the hunter's water

g. m-wana a-ki-lel-el-a ki-hembe1-child 1SA-PST-cry-APP-FV 7-knifethe child cried for a knife

The additional argument may be assigned any 0-role, except the agent 0-role.For this reason I will refer to such objects with the generic 'applied object'.

1.3 Alternative ExpressionsIn Ndendeule, only instrumental and motive have alternative

expressions with prepositional phrases as shown below.

(6) a. 13-ana 13a-ki-kayu ki-peya na ma-yanga2-child 2SA-PST-break 7-pot with 6-rockchildren broke the pot with rocks

b. Ba-lumba Ba-ki-hyem-a n-dembo u-nharu wa m-biya2-hunter 2SA-PST-hunt-FV 10-elephant 11-reason of 10-moneythe hunters hunted the elephants because of money

Swahili allows for the beneficiary to be introduced by a preposition-likestructure. The following sentences illustrate this.

4

(;

(7) a. wa-geni wa-li-m-let-e-a zawadi m-toto2-guest 2SA-PST-10A-bring-APP-FV 10present 1-childthe guests brought the child some presents

b. wa-geni wa-li-let-a zawadi kwa ajili ya m-toto2-guest 2SA-PST-bring-FV 10present for 1-childthe guests brought gifts for the child.

The applicative used in (7a) is often ambiguous in that it may be interpreted as"the child" is ultimately the recipient of the presents or the bringing was donein his stead. The paraphrase in (7b) disambiguates and means in some sensethe child is the recipient.

Other languages have these patterns too although the different objects(applied and non-applied) have different syntactic behavior cross-linguistically.Within one Bantu language also, the applied objects do not behave the sameway.

To summarize, the applicative is made by suffixing {-il-} to the verb andan additional object, the applied object. The value of the vowel in the affix is inharmony with the last vowel of the stem. The applied object may interpreted indifferent ways

A number of questions arise at this point. They form the focus of thispaper.(a) What is (-il-}?(b) Is it just an argument structure changing morpheme or is it a head of aprojection or Case assigner?(c) How do the different meanings of the applied objects arise?(d) Are all applied objects generated in the same position?(e) What are the syntactic properties of the different applied objects?

1.4 ClaimThere are three claims that I make:

(a) There are only three types of applicative whose prototypes arebenefactive, instrumental and locative.

(b) I argue on the basis of VP ellipsis that the applicative constructioninvolves a complex predicate with stacked VPs.

(c) The appled morpheme is a predicate head which selects the VP containing

the direct object.

5

Although the data for this paper is taken from Ndendeule and Swahili, I hope toshed some light on the principles determining argument structure in Bantuapplicatives.

2. Primary Object PropertiesThere is a restricted set of patterns of syntactic behavior of the applied

object. This is based on what objects may exhibit primary object propertiessuch as adjacency to the verb, object agreement, passivization andreciprocalization. With respect to constituent structure, however, allapplicative constructions exhibit the same constituents. In order to have amanageable size of data, I classify the applied objects on the basis of primaryobject properties.

Primary object properties have been used by researchers such asBresnan and Moshi (1990) as a diagnostic for determining what object is higherin the thematic hierarchy of the arguments of a sentence. Although thedifferent interpretations of applied objects have been recognized, there hasbeen no systematic study of the diagnostics to the wide range of the objects ina manner that can help identify their patterns. In this section, I show thatthere are three patterns of behavior of the applicative constructions withrespect to these diagnostics.

2.1 Object OrderIn Ndendeule and Swahili, the order of the applied and non-applied object

is more or less fixed. The object that appears immediately after the verb isconsidered primary. Arranging first the applied objects to follow the verbimmediately, the following sentences are obtained in Swahili.

2.1.1 V AO DOSwahili(8) a. Juma a-li-m-nunul-i-a rn-toto ki-tabu (Benefactive)

Juma 1SA-PST-10A-bring-APP-FV I-child 7-bookJuma brought the child a book

b. *?wa-toto wa-li-vunj-i-a ma-we ch-ungu (Instrumental)2-child 2SA-PST-break-APP 6-rock 7-pot

the children broke the pot with rocks

6

c. *wa-teja wa-li-l-i-a ofisi-ni ch-akula2-customer 2SA-PST-eat-APP-FV office-LOC 7-foodthe customers ate food in the office

d. *?wa-windaji wa-li-wind-i-a pesa ndovu2-hunter 2SA-PST-hunt-APP-FV 10money 9elephantthe hunters hunted the elephant for money

e. m-sichana a-li-sukum-i-a wa-vulana j-ongoo (Direction)1-girl 1SA-PST-push-APP-FV 2-boy 5-milipedethe girl pushed a milipede towards the boys

f. fundi a-li-kat-i-a rn-taa u-meme (Malefactive)technician 1SA-PST-cut-APP-FV 3-neighborhood 11-powerthe technician cut power to the neighborhood.

(Locative)

(Purpose)

Although the verbs are recognized as taking three arguments or twoarguments when they take the applied morpheme, in everyday use,instrumental and motive applicatives take only one of the two objects in postverbal position. Examples (8) show some problems associated with the two.The normal use of the instrumental and motive applicatives are shown in thefollowing examples from Ndendeule.

(9) a. na ngopo, pa-ki-dumul-i ni?and 10cloth, 2SA-PST-cut-APP whatand cloth, they cut with what

pa-ki-dumul-i ki-hembe2SA-PST-cut-APP 7-knifethey cut with knife

b. *pa-ki-dumul-i ngopo ki-hembe2SA-PST-cut-APP 10cloth 7-knifethey cut the cloth with a knife

c. mbona ndsmbu hi-yomw-ikiwhy I Oelephant 10SA-finish-ST+PERFwhy the elephants are finished

pa-hyem-i mbiya2SA-hunt-APP 10moncythey hunted for money

In all the responses in (9a, c) there is an empty element. In both cases it is thedirect object which is empty. Another common use of the instrumental is when

the applied object is moved to the beginning of a phrase involving anassociative and the patient/theme.

7

9

Swahili(10) a. ki-jiko cha ku-l-i-a supu

7-spoon of INF-eat-APP-FV 9sourspoon for eating soup

b. m-fuko wa ku-chukul-i-a dawa3-bag of INF-take-APP-FV lOmedicinebag for carrying medicine

These are similar to English infinitival relatives. Notice in both instrumentalapplicative constructions, only one object appears in the postverbal position.

2.1.2 V DO AOWhen the order of the objects is reversed so thai; the patient/theme

immediately follows the verb, we get different results from the V AO DO.

Swahili(11) a. *Juma a-li-m-nunul-i-a ki-tabu m-toto

Juma 1SA-PST-10A-buy-APP-FV 7-book 1-childJuma bought the child a book

b. wa-toto wa-li-vunj-i-a ch-ungu ma-we (Instrumental)2-child 2SA-PST-break-APP 7-pot 6-rock

the children broke the pot with rocks

c. wa-teja wa-li-l-i-a ch-akula ofisi-ni (Locative)2-customer 2SA-PST-eat-APP-FV 7-food office-LOCthe customers ate food in the office

(Benefactive)

d. *wa-windaji wa-li-wind-i-a ndovu pesa (Purpose)2-hunter 2SA-PST-hunt-APP-FV 9elephant 10moneythe hunters hunted the elephant for money

e. *m-sichana a-li-sukum-i-a j-ongoo wa-vulana (Direction)1-girl 1SA-PST-push-APP-FV 5-milipede 2-boythe girl pushed a milipede towards the boys

f. *fundi a-li-kat-i-a u-meme m-taa (Malefactive)technician 1SA-PST-cut-APP-FV 11-power 3-neighborhoodthe technician cut power to the neighborhood.

If we leave out (11b, d), we find that we get exactly the opposite results of whatwas obtained when the applied object were all lined up immediately followingthe verb. Therefore, if adjacency to the verb is considered a feature of primaryobjects, then beneficiaries, goals and maleficiaries are all primary objects. Incontrast, in instrumental, motive and locative applicatives, the theme/patient

8

1 0

is the primary object. These results are also found when object agreement isconsidered.

2.2. Object agreementThere are also restrictions to objects with respect to object marking. In

Ndendeule and Swahili only one of the objects may be marked on the verb. Theobject that is closest to the verb is the one that is marked. In the followingexamples, I present sentences first with the applied object adjacent to the verband then another set in which the direct object is closest to the verb.

2.2.1 Applied object marked in the verbSome applied objects may be marked in the verb and others may not be

marked as the following examples from Swahili show.

(12) a. Juma a-li-m-nunul-i-a m-toto ki-tabuJuma 1SA-PST-10A-bring-APP-FV I-child 7-bookJuma brought the child a book

b. *wa-toto wa-li-ya-vunj-i-a ma-we ch-uagu (Instrumental)2SA-PST-60A-break-APP 6-rock 7-pot

the children broke the pot with rocks

c. *wa-teja wa-li-pa-l-i-a ch-akula ofisi-ni (Locative)2-customer 2SA-PST-160A-eat-APP-FV 7-food 9office-LOCthe customers ate food in the office

(Benefactive)

d. *wa-windaji wa-li-zi-wind-i-a ndovu pesa (Purpose)2-hunter 2SA-PST-I00A-hunt-APP-FV 9elephant 10moneythe hunters hunted the elephant for money

e. rn-sichana a-li-wa-sukum-i-a wa-vulana j-ongoo (Direction)1-girl 1SA-PST-20A-push-APP-IN 2-boy 5-milipedethe girl pushed a milipede towards the boys

f. fundi a-li-i-kat-i-a mi-taa u-meme (Malefactive)technician 1SA-PST-40A-cut-APP-FV 4-neighborhood 11-powerthe technician cut power to the neighborhoods.

The results of this diagnostic are the same as those for word order consideredabove. The applied objects, instruments (12b), location (12c) and motive (12d),

may not be marked in the verb. Beneficiary (12a), maleficiary (120 and goal(12e) may be marked on the verb. Changing the order of the objects gives us

the opposite of what we have seen.

9

1 1

2.2.2 Direct Object marked in the verbWhen the object marking is reversed so that the direct object precedes

the applied object, different results are obtained. This is shown in the followingexamples from Swahili.

(13) a. *Juma m-toto ki-tabuJumii 1SA-PST-7A-bring-APP-FV 1-child 7-bookJuma brought the child a book

(Benefactive)

b. ?wa-toto wa-li-ki-vunj-i-a ch-ungu ma-we (Instrumental)2-child 2SA-PST-70A-break-APP 6-rock 7-pot

the children broke the pot with rocks

c. wa-teja wa-li-ki-l-i-a ch-akula ofisi-ni2-customer 2SA-PST-70A-eat-APP-FV 7-food 9office-LOCthe customers ate food in the office

d. ?wa-windaji wa-li-wa-wind-i-a ndovu pesa2-hunter 2SA-PST-20A-hunt-APP-FV 2elephant 10moneythe hunters hunted the elephant for money

e. *m-sichana a-li-rn-sukum-i-a wa-vulana jongoo1-girl 1SA-PST-10A-push-APP-FV 2-boy 1-milipedethe girl pushed a milipede towards the boys

(Locative)

(Motive)

(Goal)

f. *fundi a-li-u-kat-i-a mi-taa u-meme (Malefactive)technician 1SA-PST-110A-cut-APP-FV 4-neighborhood 11 -powerthe technician cut power to the neighborhoods.

It seems only the direct object in the locative can be object-marked withoutproblems. The appearance of both objects in postverbal position ininstrumental and motive applicatives has some problems. With thisdiagnostic, it is possible to improve the instrumental and motive applicativesentences if the direct object is marked on the verb but only the applied objectis assigned phonetic content. Another significant improvement in instrumentaland purpose when the direct object is [+human].

From this I conclude that the patient/theme argument in instrumentaland motive applicative may be marked in the verb. These results are actuallyreciprocal to those obtained under 2.2.1. The results obtained so far on objectorder and object marking pattern the same way as in passivization.

2.3. PassivizationIt is assumed that primary objects may be subjects of passives. I

compare the applied objects with the direct objects.

100

2.3.1 Passivization of the Applied ObjectIn passive constructions, the object moves to the subject position. Here,

I move the applied object first as the following examples in Swahili show. Themoved applied object is underlined.

(14) a. m-toto a-li-nunul-i-w-a ki-tabu (Benefactive)1-child 1SA-PST-10A-bring-APP-PASS-FV 7-book

The child was bought

b. ?ma-we ya-li-vunj-i-w-a ch-ungu (Instrumental)6-rock 6SA-PST-break-APP-PASS-FV 7-pot

the rocks were broken pots

c. ofisi-ni pa-li--1-i-w-a ch-akula (Locative)office-LOC 16SA-PST-eat-APP-PASS-FV 7-foodin the office was eaten food

d. *pesa zi-li-wind-i-w-a ndovu (Purpose)10money 10SA-PST--hunt-APP-PASS-FV 9elephantmoney was hunted an elephant for

e. wa-vulana wa-li-sukum-i-w-a j-ongoo (Goal)2-boy 2SA-PST-push-APP-PASS-FV 5-milipedethe boys were pushed a milipede to

f. mi-taa a-li-kat-i-w-a u-meme (Malefactive)4-neighborhood 1SA-PST-cut-APP-PASS-FV 11-powerneighborhood were cut power for (i.e. at the expense of)

Passivization of the instrumental object in this test shows a very doubtfulcase. The purpose phrase in subject position does not give a grammaticalsentence. The rest of the applied objects may move to the subject position ofthe passive. This includes the locative argument as seen in (14c). The resultsare reversed when the moved phrase is the direct object.

2.3.2 Passivization of the Direct ObjectsThe direct object is moved to the subject position. The sentences have

the same elements as those in (14) above.

(15) a. *ki-tabu ki-li-nunul-i-w-a m-toto7-book 7SA-PST-buy-APP-PASS-FV 1-childthe book was bought the child

11

1 :J

b. ?ch-ungu ki-li-vunj-i-w-a ma-we7-pot 7SA-PST-break-APP-PASS-FV 6-rockthe pot was broken with rocks

C. ch-akula ki-li-l-i-w-a ofisi-ni7-food 7SA-PST-eat-APP-PASS-FV 9office-LOCthe food was eaten in the office

d. ?ndovu wa-li-wind-i-w-a pesa2elephant 2SA-PST-hunt-APP-PASS-FV 10moneyelephants were hunted for money

e. *jongoo a-li-sukum-i-w-a wa-vulanaI milipede 1SA-PST-push-APP-PASS-FV 2-boythe milipede was pushed towards the boys

f. *u-meme u-li-kat-i-w-a mi-taa11-electricity 1 1SA-PST-cui-APP-PASS-FV 4-neighborhoodpower was cut for the neighborhoods

This time in (15) we have both instrumental and motive applicative passivesas marginally grammatical. The direct object in locative applicative can be the

subject of the passive without any problem. The direct object in benefactives,goal and malefactives may not be the subject of the passive.

2.4 ReciprocalizationThe reciprocal in both Swahili and Ndendeule consists of the subject and

a verb to which the reciprocal -an- is suffixed. Example (16) below illustrates a

simple reciprocal in Ndendeule.

(16) a. pa-na p-i-n-liy-a tati yw-ape2-child 2SA-PRT-10A-insult-FV father 1-theirthe children are insulting their father

b. pa-na p-i-liy-an-a2-child 2SA-PRT-insult-REC-FVthe children are insulting each other

There is no object in (16). There is only a plural subject. In applicativeconstructions, one object reciprocalizes and therefore leaves a gap in thepostverbal position. The other object which is not reciprocalized is fully realized

as the following example shows.

12

1 4

Swahili(17) a. wa-geni wa-li-nunul-i-an-a zawadi

2-guest 2SA-PST-buy-APP-REC-FV 10presentguests bought presents for each other

b. wa-geni wa-li-pig-an-i-a zawadi2-guest 2SA-PST-hit-REC-APP-FV lOpresentguests hit each other for presents

These two examples illustrate three things. One, the order of the suffixes canbe APP+REC as shown in (17a), or it may be REC+APP as in (17b). There isalso a different in meaning involved. Secondly, the missing/reciprocalized objectcan be the applied object as in (17a) or it may be the direct object as in (17b).Thirdly, the order of the two suffixes correlates with which argument will bereciprocalized.

Objects in applicative constructions do not reciprocalize freely. Someapplied objects do reciprocalize and some do not reciprocalize. The pattern,however, repeats what I have already shown regarding adjacency.to the verb,object agreement and passivization. I provide test examples in which theapplied object reciprocalizes and then reciprocalize direct objects.

2.4.1 Reciprocalize the Applied ObjectIn this set of examples, the reciprocalized (and therefore the missing)

object is the applied object.

Ndendeule(18) a. pa-ki-kom-81-an-a mbuhi

2SA-PST-kill-APP-REC-FV lOgoatthey killed goats for each other

b. *pa-ki-hip-il-an-a n-dyango2SA-PST-close-APP-REC 3-door

they used each other to block the door

c. *pa-lci-la-El-an-a mi-yonda2SA-PST-leave-APP-REC-FV 4-farmthey left the farms because of each other

d. pa-chongolo pa-ki-kang-il-an-a ma-yongo102-boy 2SA-PST-push-APP-REC-FV 6-milipedeboys pushed milipedes to each other

e. pakiyomolslana mbiya2SA-PST-finish-APP-REC-FV lOmoneythey finished each other's money

13

In this set of examples, I have used the APP+REC order. In this order, thebeneficiary, maleficiary and goal may reciprocalize as (18a,d,e). Theinstrument and motive may not reciprocalize as seen in (18b,c). I could notmake a sentence in which the locative is reciprocalized.

2.4.2 Reciprocalizing the Direct ObjectWith the same order, APP + REC, the missing object (i.e. the

reciprocalized object) is the direct object.

Ndendeule(19) a. *3a-ki-kom-e1-an-a rnbuhi

2SA-PST-kill-APP-REC-FV iOgoatthey killed each other for the (benefit of) goats

b. *3a-ki-tem-e1-an-a ki-hembe2SA-PST-cut-APP-REC-FV 7-knife

they cut each other with a knife

c. *pa-ki-tem-el-an-a ku-ki-hinja2SA-PST-cut-APP-REC-FV 17-7-kitchenthey cut each other in the kitchen

d. *pa-ki-1ek-e1-an-a mi-yonda2SA-PST-leave-REC-APP-FV 4-farmthey left each other because of farms

e. *pa-chongolo pa-ki-kang-il-an-a li-yongolo2-boy 2SA-PST-push-APP-REC-FV 5-milipede

boys pushed each other to the miiipede

In order to successfully reciprocalize the direct object, the order of the twosuffixes must be REC + APP.

Ndendeule(20) a. pa-ki-kom-an-i mbuhi

2SA-PST-kill-REC-APP lOgoatthey killed each other for the (because of) goats

b. pa-ki-tsm-an-i ki-hembe2SA-PST-cut-REC-APP 7-knifethey cut each other with a knife

c. pa-ki-tem-an-i ku-ki-hinja2SA-PST-cut-REC-APP I 7-7-kitchenthey cut each other in the kitchen

14

d. pa-ki-lck-an-i mi-yonda2SA-PST-leave-REC-APP 4-farmthey left each other because of farms

e. *pa-ki-ksm-an-i 13a-hana pa-pc2SA-PST-call-REC-APP 2-wife 2-theirthey called each other for their wives

With the order REC+APP the direct object is reciprocalized in instrumental,motive and locative applicatives (20a-d). The direct object in benefactiveapplicatives may not be reciprocalized when the order of the suffixes isREC+APP as shown in (20e).

Once more the applied object may be reciprocalized if it is a beneficiary,goal or maleficiary. The order of the morphemes in such cases is APP+REC. Ifthis order of morphemes is maintained no sentence of instrumental, motive orlocative comes out grammatical (see 18). However, if the order of themorphemes is reversed so that we have REC+APP as seen in (20), then we getinstrumental, locative and motive readings. The reciprocalized argument is notthe applied object. The reciprocalized argument is the direct object.

3. A Typology of Applicatives

3.1 Previous TypologyFew studies have dealt with organizing the diverse applied objects into

types. The few include Baker (1988b, 1990, 1992), Alsina and Mchombo(1988), Marantz (1993) and Hoffman (1991). Even when tyology is discussed,most of the applied objects are not clearly discussed with respec to otherapplicatives they pattern with. I will discuss briefly typology as discussed bythese in so far as they have a bearing in my analysis of the applicatives.

3.1.1 Baker's TypologyBaker (1998b, 1990, 1992) identifies two types of applicatives:

benefactives and instrumentals. He further suggests that the locativeapplicatives belong to the benefactive type. There are at least four ways inwhich benefactive applicative differs from instrumental applicative.

15

(a) Primary object properties: In the language he studied, Chichewa, only thebeneficiary exhibits primary object properties in benefactive applicatives. Ininstrumental applicatives either object may exhibit primary object properties.

(b) Theta-theory: Benefactives, like locatives, are arguments of prepositions.They are 0-marked by the preposition, in this case, the applied affix. Ininstrumental applicatives, both the instrument and the theme/patient arearguments of the verb. They are both 6-marked by the verb.

(c) The structure: The structure of the two applicatives incorporating the 0-theory features is as follows.

(21) a. Benefactive

NP VP

Mavuto V PP NPI

mold-APPi P 111P waterpot

ti chibf

b. Instrumental

NP VP

MavutoV NP NP

mold-APP knife waterpot

In either case, the VP has a ternary branching structure. The beneficiary in(21a) is a complement of a preposition which has moved out to incorporate intothe verb. In instrumental applicatives, both the instrument and the theme arecomplements of the applicative verb.

(d) Case Theory: From the structure, differences in Case between the twotypes are derived. Following Chomsky (1986) and Belletti (1988), Baker(1988a, 1990, 1992) assumes structural Case is assigned under governmentat S-structure while inherent Case is assigned under government and 0-marking at D-structure. Structural Case is assigned to NP adjacent to thecase assigner. This adjacency is not necessary in inherent Case. Eachtransitive verb can assign both structural case and inherent Case. Inbenefactive qpplicatives, the verb with its applicative suffix, assignsstructural Case to the beneficiary. Having moved out of the PP, thepreposition cannot assign Case, neither can its trace. The theme receivesinherent Case. Such a trace is not found in instrumental applicatives. Here,

16

18

the verb with the applicative suffix assigns structural Case to one object andinherent Case to the other freely.

This account is unclear on many points. The structures presented aboveare based on stipulation rather than evidence from Chichewa. In particular,the typology is based on classifying the applicative morpheme, in one case as apreposition and in another as a non-incorporated suffix. There is no evidencesupporting this classification. This is carried further when beneficiaries arecompared to locatives as arguments of prepositional phrases. With respect toprimary object properties, locatives behave in many ways like instrumentalsas seen in Section 2 above and as Alsina and Mchombo (1990) point out. Theternary branches of the VPs present a particular empirical problem as noevidence is presented to support the three way branching. This structuremakes a particular prediction that processes affecting the VP will always showthe effects on both objects in all applicative constructions. An example of VPellipsis will show that this is not the case.

Ndendeule(22) a. n-haru tati a-ki-dumul-i n-kota ki-hembe,

3-reason father 1-PST-cut-APP 3-sugar cane 7-knifesince the father cut the sugar cane with a knife,

na mw-ana a-ki-dumul-I n-kota ki-hembe helaheand 1-child 1SA-PST-cut-APP 3-sugar cane 7-knife tooand the child cut the sugar cane with a knife too

b. na mw-ana a-ki-dumul4 n-ksta- ki-hembe helaheand 1-child 1SA-PST-cut-APP 7-knife tooand the child cut the sugar cane with a knife too

c. * na mw-ana a-ki-dumul4 n-k3ta k=i-hembe helaheand I-child 1SA-PST-cut-APP 3-sugar cane tuoand the child cut the sugar cane with too

d. na mw-ana a-ki-dumul-I n-k9ta-- --ki-hernbe helaheand 1-child I SA-PST-cut-APP tooand the child did too

I have shown in these examples that VP ellipsis can affect one object only as in(22b) and can affect both (22d). It shows also that the applied object may notbe affected alone. This suggests the structure presented in (21b) is incorrect.With it go all the stipulations about 0-role assignments.

17

Alsina and Mchombo (1990) point out similarities between instrumentaland locative applicatives with respect to primary object properties. As can beseen in Section 2 above also, in locative applicatives, just as in instrumentalapplicatives, the theme is the object adjacent to the verb, may cliticize andpassivize. But in addition, the locative may passivize. This calls for looking atthe locative applicatives on their own.

The most important point from Baker's typology, however, is the ideathat the typology is structurally motivated. This idea is also found in Marantz'sdiscussion of the asymmetry between the benefactive applicatives andinstrumental applicatives.

3.1.2 Marantz (1993)Marantz (1993) following Barss and Lasnik (1986) and Larson (1988)

analyses the applicative construction as a complex predicate. The beneficiaryis outside the the event structure which contains the direct object. For thisreason, the beneficiary is located in in the higher VP. The instrument and thelocative are in the event structure containing the direct object. The exactposition of the two objects is not dictated by the event structure. That is whyeither object in instrumental applicative may be higher.

Marantz's analysis is mainly based on asymmetries in c-commandrelations. As Marantz (1993) points out, c-command tests (as those in Barssand Lasnik (1986)) for the applicative constructions are not possible exceptthe quantified noun phrase (QNP) -pronoun relation. There is an asymmmetrybetween benefactive applicatives and instrumental applicative with respect toQNP-pronoun relations. The following examples from Ndendeule have thebound pronoun italicized in benefactive applicative.

(23) a. na-ki-m-pek-e kila mu-ndu hundi hy-akeI-PST-10A-put-APP-FV each 1-person lOcheck 10-hisI put aside for each person his checks

b. 'na-ki-m-pek-e hundi hy-ake kila mu-ndu1-PST-10A-put-APP-FV lOcheck 10-his each 10-workerI put for each worker his checks

The possessive pronoun in (23a) is bound by a c-commanding QNP "each". Inbenefactive applicatives, only the beneficiary may bind into the theme/patient.The theme/patient may not bind into the beneficiary. The conclusion here isunambiguous that the beneficiary is higher than the theme/patient.

18

The following sentences from Ndendeule show the QNP-pronoun relationin instrumental applicatives.

(24) a. a-ki-dindul-i kila n-dyango pungulu y-akeISA-PST-open-APP each 3-door 9key 9-itsshe opened each door with it's key

b. *a-ki-dindul-i n-dyango w-ake kila punguluI SA-PST-open-APP 3-door 3-its each 9keyshe opened its door with each key

c. a-ki-dindul-i kila pungulu n-dyango w-ake1SA-PST-open-APP each 9key 3-door 3-itshe opened with each key its door

d. *a-ki-dindu1-i pungulu y-ake kila n-dyangoISA-PST-open-APP 9key 9-its each 3-doorhe started with its key each car

In (24a) the QNP direct object binds into the instrument. In this case, theinstrument must be in the scope of the patient. The patient c-commands theinstrument. If the order of the objects is reversed such that the instrumentprecedes the patient, the quantified instrument binds into the patient (24c). Inthis case, the patient is in the scope of the instrument. The instrument c-commands the patient.

The locative applicatives also have a different pattern from thebenefactive and instrumental applicatives. This can be seen in the Ndendeulesentences in (25).

(25) a. pa-ki-m-pol-e kila mu-ndu pa-kaya p-ake2SA-PST-10A-teach-APP each 1-person 16-9horne I6-histhey taught each person at her home

b. *13a-ki-13o1-e kila pa-kaya mu-ndu w-ake2SA-PST-teach-APP each 16-9home I-person I-itsthey taught at each home its person

Only the patient may bind into the locative phrase (25a). The reverse is notpossible (25b). This means only the patient can c-command the locativephrase. There is a constraint for locative phrases because they do not appearin QNP.

The results of this tests are summarized in the following table.

19

(26) QNP- ronoun relations

BENEFACTIVE INSTRUMENTAL LOCATIVEApplied Object binds intoDirect Object 4 4

,

*

Direct Object binds intoApplied Object

* 4 4

In instrumental applicatives either object may bind into the other. Marantzconcludes because either may occupy either of the two object positions.

Example (22) above, however, suggests Marantz's analysis is incorrectbecause VP deletion does not target a constituent with the instrument butwhich does not contain the direct object. The DO-AO order must be derived bysome movement to some quantifier position which allows the direct object tobind into the instrument.

3.2 Typology Based on Object PropertiesThe results of the four diagnostic tests, adjacency to the verb, object

agreement, pasivization and reciprocalization consistently show that thebeneficiary, the goal and the maleficiary are applied objects which exhibitprimary object properties. The direct objects which appear with these appliedobjects do not exhibit primary object properties. On the other hand, theinstrument, motive and locative do not exhibit primary object properties. Theirdirect objects do exhibit primary object properties. Although the locativeargument shows almost all features of a non primary object, it may passivize.

These results may summarized in a table in which I have arrangedaccording to their behavior.

(27) A Checklist of Primary Object Properties1 BENEF, GOAL MALEF. INSTRUM MOTIVE LOCAT.

OBJ. ORDER AO DO4 4 4

* * *

DO AO :I:? ? 4 \I

OBJ. AGR. APP. 0I3J.4 \ 1 4 ? *

DIR. OBJ. *?4 q 4

PASS. SUBJ. APP. OBJ. ? 4DIR. OBJ * * *

RECIPROC. APP. 011.1. *

DIR. OBJ. * * *

Key to JudgmentsGrammaticalMarginalUngrammatical

From the foregoing, three types of applicatives emerge:

(a) Benefactive, goal and malefactive. I will consider the benefactive as theprototype for this group;

(b) Instrumental and motive. I will take the instrument to be the prototype.(c) Locative as a unique applicative.

This typology goes against Baker's (1988a,1990, 1992) and provides detailsnot presented before in Alsina and Mchombo (1990).

4. VP EllipsisVP ellipsis is a good diagnostic for the internal structure of the

applicative predicate. In both Swahili and Ndendeule, VP ellipsis eliminates thecomplement of the verb. The verb is not eliminated as can be seen in thefollowing examples.

Swahili(28) mw-alimu a-li-nunu-a ki-tabu cha Chomsky

1-teacher 1SA-PST-buy-FV 7-book of Chomskythe teacher bought Chomsky's book

na wa-nafunzi wa-li-nunu-a pia.and 2-student 2SA-PST-buy-FV tooand students did too

The verb is the tense carrier. The verb has moved out of the VP in order tosupport tense. This is unlike VP ellipsis in English where tense moves to the Vin non-elliptical structure. In VP ellipsis, the verb deletes together with thecomplement and if there is no auxiliary, tense is supported by an inserted do asthe following example shows with its structure in (29b).

(29) a. John bought a house

b. IP

DP I

John I VP tarpt of VP deletionI 7\

-ed V

V DP1

buy a house

In Swahili, however, the verb does not disappear.

(30) a. Juma a-li-nunu-a ki-tabu cha ChornskyJuma I SA-PST-buy-FV 7-book of ChomskyJuma bought Chomsky's book

b. IP

DP I

Juma 17VP target of VP deletionV\alinunua V

,rvNV DP

t kitabu cha Chornsky

An alternative to the NT ellipsis analysis is to regard the gap in thesecond clause of (28) as a null object. This analysis is not correct for at leastfour reasons. The first reason is derived from facts about object clitics. AsBergvall (1986) showed, in Kikuyu and Swahili an null object can only be foundin construction in which the verb has an object clitic. This means that for theBantu languages, like Romance languages, the null object shares 0-features(gender and number) with a coindexed object (laic. In example (28) there is noobject clitic and therefore I conclude the gap lacks the nominal features.Therefore it is not an object. gap.

Secondly, given that null objects need a coindexed object ciitic, a gapinvolving two objects shows all the more tha t it is not pronominal. BothNdendeule and Swahili do not have double clitics. Therefore when we see a gapfor double object constructions such as datives and applicatives, we shouldconclude that it is devoid of the 0-features of the two objects.

22

2 4

The third argument against a null object analysis is that thedisappearance may affect elements which are not objects.

Swahili(31)a. mama a-li-tak-a ku-m-nunul-i-a m-toto vi-atu

mother I SA-PST-want-FV INF- I OA-buy-APP-FV f-child 8-shoemother wanted to buy the chiid shoes

na baba a-li-tak-a --- pia9nd father 1SA-PST-want-FV tooand father wanted too

b. wa-limu wa-li-end-a shamba-ni na wa-nafunzi wa-li-end-a --- pia2-teacher 2SA-PST-go-FV 5farm-LOC and 2-student 2SA-PST-go-FV tooteachers went to the farm and their students did too.

In (31a) the complement of taka (want) is an infinitival clause. Thedisappearance of this infinitival complement in the second clause cannot beattributed to a null pronoun. Similarly, no null pronoun can be posited for thelocative complement of the second clause of (31b) because the complement isnot an object.

The fourth argument is that idiom chunks behave in the sameway. Idioms appear after specific verbs. One restriction of idiom chunks is thatthey may not be pronominalized. I give examples from English.

(32) a. The president pays lip service to the people

b. *The president pays it to the people

c. The ruling party took advantage of the confusion in the opposition

d. *The ruling party took it of the confusion in the opposition

Sentence (32a,c) have the object idioms which are italicized. Following themare pronominalized forms of the object idioms in (32b,c) Both areungrammatical. If we cliticize the object found in the idiom, we getungrammatical sentences as shown in the following examples.

Swahili(33) a. dada a-lipig-a simu

sister I SA-PST-hit-IN 9-telephonesister called

b. *dada a-li-i-pig-asister I SA-PST-90A-hit-FVsister called

23

c. Mumbi a-li-kul-a ki-apoMumbi 1SA-PST-eat-FV 7-oathMumbi took the oath

d. *Mumbi a-li-ki-l-aMumbi 1SA-PS'I.-70A-eat-FVMumbi took the oath

I have indicated the object idiom by italics in (33a,c). A word for wordtranslation of the VP in (33a) is hit the phone. Pronominalization of the phonein (33b) results in an ungrammatical structure. The VP in (33c) translates aseat the oath. Pronominalization of "the oath" is bad.

Although the object idiom cannot be pronominalized, it does disappearunder VP ellipsis.

Swahili(34) a. dada a-li-pig-a sirnu na mama a-li-pig-a pia

sister 1SA-PST-hit-FV 9-telephone and mother 1SA-PST-hit-FV --- alsosister called and mother did too

b. Mumbi a-li-kul-a ki-apo na Njorogc a-li-kul-a piaMumbi 1SA-PST-eat-FV 7-oath and Njorore 1SA-PST-eat-FV alsoMumbi took the oath and Njorcr.- did too

In both these sentences, the objects have disappeared. Therefore,pronominalization and disappearance of the object under VP ellipsis are twodifferent things. In (34) we have two examples of VP ellipsis which are notconsistent with pronominalization.

I have argue that the gap found in the complement of the verbs isactually a case of VP ellipsis. The surface realization of the verb is derivedfrom the fact that the verb raises to an Infl. node. This movement of the verb issimilar to what is found in Romance languages and in Japanese, Korean andChinese (Otani and Whitman, 1991).

4.2 VP Ellipsis in Applicative ConstructionsVP ellipsis makes an interesting prediction regarding the the structure

of applicative constructions. If both objects are in the same simple VP, thenboth should disappear under VP ellipsis. However, if they are in different VPs,one should be able to disappear without the other. Further, f one of the VPs is

a constituent of the other, it should be impossible for one of the objects to

24

2 6

disappear without the other. I use this diagnostic with the three types ofapplicatives, benefactive, instrumental and locative.

4.2.1. VP Ellipsis in Benefactive ApplicativeThe two objects in benefactive applicatives are the beneficiary (or goal,

maleficiary, etc) and the direct object. Evidence from benefactive applicativessuggests that there are two VPs in the applicative construction. The followingset of examples are from Ndendeule.

(35) a. n-gheni a-ki-hsm214 shuli hi-tepo1-guest 1SA-PST-buy-APP 9school 8-chairthe guest bought the school chairs

na hokolo a-ki-hemE1-g shuli hi-tebo helahéand grandpa I SA-PST-buy-APP 9school 8-chair tooand grandpa bought the school chair

b. na hokolo a-ki-hems1-e shuli hi-tebo helahéand grandpa I SA-PST-buy-APP 9school 8-chair tooand grandpa bought the school chairs

c. *na hokolo a-ki-fisms1-g shuli hi-tebo he1ahdand grandpa I SA-PST-buy-APP 9school 8-chair tooand grandpa bought the school chairs too

d. na hokolo a-ki-h8m81-f; shuli hi-tebo helahéand grandpa 1SA-PST-buy-APP 9school 8-chair tooand grandpa did too

In (35b), only the theme disappears. The beneficiary is a remnant. Thissentence is grammatical. In (35c), however, the beneficiary is missing leavingbehind the theme. This is ungrammatical. In (35d), both the theme and thebeneficiary disappear. This sentence is grammatical.

Two important conclusions are derived from these results. One is thatthe benefactive applicative construction consists of two VPs because it ispossible for the theme to disappear leaving behind the beneficiary. Secondly,the beneficiary may only be disappear together with the theme. This meansthe beneficiary forms a larger constituent VP together with the VP whichcontains the theme. That is to say the theme is the innermost object, object of

25

27

the subordinate VP. A simple tree diagram such as the following shows thestacked VP structure and how the VP ellipsis results may be derived.

(36) a. hokolo shuli hi-tebo he1ah6grandpa 1SA-PST-buy-APP 9school 8-chair toograndpa bought the school chair

b. IP

DP l'I

hokolo I VP, target of VP deletion

I

akihemsle shuli V'

V VP1 target of VP deletion

V'

V DP

t hite po

Deletion of VP1 does not affect the beneficiary. Only the theme disappears.Deletion of VP2, however, causes both the theme and the beneficiary todisappear. The beneficiary, therefore, must be in the specifier of the VP2. Thisposition of the beneficiary is the same as that of the instrument ininstrumental applicative to which I now turn.

4.2.2. VP Ellipsis in Instrumental ApplicativesUnlike benefactive applicatives in which the applied object (the

benefactive) exhibits primary object properties, in instrumental applicatives,the theme/patient exhibit primary object properties. The pattern of VP ellipsisobtained when examining the benefactive applicative is the same for theinstrumental applicatives as the following set of examples show.

(37) a. n-haru tati a-ki-dumul-i n-kota ki-hembe,3-reason father 1-PST-cut-APP 3-sugar cane 7-knifesince the father cut the sugar cane with a knife,

na mw-ana a-ki-durnul-i n-kota ki-hembe helaheand 1-child 1SA-PST-cut-APP 3-sugar cane 7-knife tooand the child cut the sugar cane with a knife too

26

28

b. na mw-ana a-ki-dumul-i n-keta ki-hembe helaheand 1-child 1SA-PST-cut-APP 7-knife tooand the child cut the sugar cane with a knife too

c. * na rnw-ana a-ki-dumul-i n-kota ki-henibe helaheand 1-child 1SA-PST-cut-APP 3-sugar cane tooand the child cut the sugar cane with too

d. na mw-ana a-ki-dumul-i n-keta ki hembe helaheand 1-child 1SA-PST-cut-APP tooand the child did too

Like the benefactive applicative, the instrumental applicative allows for theVP containing the patient to delete while the instrument is a remnant (37b). Itis also possible to delete the VP containing the instrument together with theVP containing the patient (37d). However, there is no VP such that it containsonly the instrument without the patient. That is why (37c) is ungrammatical.Therefore, the structure is agair: of stacked VPs like that of benefactiveapplicative. The [Spec, VP2] position is the same position at which theinstrument is generated.

This raises an interesting prospect of generating all the applied objectsfrom the same structural position. In order to determine this, I now apply thediagnostic to locative applicatives.

4.2.3 VP Ellipsis in Locative ApplicativesIn locative applicatives, the theme/patient exhibit primary object

properties. In a way this is similar to instrumental applicatives where thetheme/patient also exhibits primary object properties. The behavior of locativeapplicatives in VP ellipsis is the same as that found in benefactive andinstrumental applicatives. This is illustrated in the following set of examples.

(38) a. n-gheni a-ki-1e1-6 nhElE mu-nyumba1-guest 1SA-PST-eat-A PP rice 18-9-housethe guest ate rice in the house

na hokolo n-helc mu-nyumba helaheand grandpa 1SA-PST-eat-APP 3-rice 18-9house alsoand grandpa ate rice in the house too

b. na hokolo a n-heic mu-tiyumba helaheand grandpa 1SA-PST-eat-APP 18-9house alsoand grandpa did -- in the house too

27

29

c. * na hokolo a-ki-le1-6 n-hcle ni-u---nyurnb-a helaheand grandpa ISA-PST-cat-APP 3-rice alsoand grandpa ate rice in too

d. na hokolo a-ki-le1-6 n hele mu nyumha helaheand grandpa 1SA-PST-eat-APP alsoand grandpa did too

In locative applicative sentences, just as in benefactive and instrumentalapplicatives, the theme/patient appears in the lowest VP. When this VP isdeleted, the remnant is a VP in which we find the applied object, in this case thelocative (38b). If VP deletion targets the higher VP which contains the locative,the result is the lower-VP which is a constituent of the higher VP disappearswith it (38d). There is no way VP ellipsis can target the locative argumentalone as (38c) shows.

4.2.4 A Unified Account of VP Ellipsis in Applicative ConstructionsThe foregoing discussion has shown that although the different appled

objects behave differently with respect to primary object properties, they allbehave the same way with respect to VP ellipsis. The following tablesummarizes the results of this diagnostic. Therefore, under the Benefactivecolumn. the AO is the beneficiary. Under instrumental, it is the instrumentand under locative the AO is a locative phrase. In this table, the objects arearranged in their preferred order.

(39) VP deletion

BENEFACTIVE INSTRUMENTAL LOCATIVE

Sub' V AO DO Sub' V DO AO Sub' V DO AO

Subj V AO DO Subj V DO AO Subj V DO AO

*Subj V AO DO *Subj V DO AO *Subj V DO AC)

Subj V AODO Subj V DOA0 Subj V DOA0

The different object order notwithstanding, direct may be targeted at theexclusion of the applied object. The applied object may only be affected whenboth objects are involved. In all three types of applicatives, the theme/patientis in the lower VP and the applied object is in the higher VP.

28

3 0

Three crucial conclusions are arrived at. One is that the Verb raises toan Infl node as also demonstrated in the simple predicate sentences. Secondly,

the applied object is not within the minimal VP. The direct object(theme/patient) is in the minimal VP. Thirdly, the VPs are stacked.

4.3 Universal Structure of the ApplicativeBefore I present the proposed universal structure of applicative

constructions, I need to state that I assume a strictly local relationshipbetween the verb and the complement (following Chomsky, 1993),Furthermore, I assume Baker's Uniformity of Theta Assignment Hypothesis(UTAH) which states that:

(40) Identical thematic relationships between items are represented byidentical structural relationships between those items at the level of D-structure (Baker, 1988b p.46).

Therefore themes/patients must be in a local relationship with V. The appliedobject which is licensed by the applicative morpheme must be in a localrelationship.

Recall that the theme is in the minimal VP. The V is generated in thisminimal VP. This minimal VP is a complement of a higher V. I will tentativelypropose that the higher V is the applicative morpheme. The other availableposition for this higher projection is the spec position. If we assume that theapplied object is generated in this position, we maintain the local relationshipbetween the head (the applicative morpheme) and the argument it licenses(the applied object). The first approximation of the universal applicativeconstruction can be presented this way.

(41) IP

bp

Subject I VP2 target of VP deletion

I

AO V'

V VP1 ( target of VP deletion

DP V'

DO V

29

31

If VP ellipsis targets the lowest VP, only the theme/patient disappears. If ittargets the higher VP, then both objects disappear.

This structure is in many ways similar to that of Marantz (1993). Hebased his proposal on c-command relations determined by quantified nounphrase-pronoun relations following Barss and Lasnik (1986) and Larson (1988)complex predicate analysis.

5. ConclusionTo summarize the a -guments of this paper: Based on primary object

properties, there are throe types of applicatives whose prototypes arebenefactive, instrumental and locative. Using VP ellipsis as a diagnostic forconstituent structure, I arrived at the conclusion that there are two VPsinvolved. One VP which contains the theme/patient is embedded in another VPwhich contains the applied object. I conclude also that the applicativemorpheme is a predicate head which selects the minimal VP and the appliedobject.

The crucial question now is how are the arguments mapped onto thesentence structure. I am inclined to consider the meaning differences of thevarious types of applicatives to be a factor of structural differences (Hale andKeyser, 1993) which go beyond the universal structure of the applicativeconstruction proposed in (41). Consider for example instrumentals. There aretwo types of instruments.

Swahili(42) a. m-toto a-li-l-i-a wali ki-jiko (aide)

1-child 1 SA-PST-eat-APP-FV 1 1-rice 7-spoonchild ate rice with a spoon

b. m-toto a-li-kat-i-a mu-wa ki-su (tool)1-child 1SA-PST-cut-APP-FV 3-sugar cane 7-knifechild cut sugar cane with a knife

As in English, the two types of instruments differ in Swahili too. The followingtwo sentences illustrates such a difference.

30

3"'

Swahili(43) a. ki-su ki-na-kat-a mi-wa vi-zuri

7-knife 7SA-PRT-cut-FV 4-sugar cane Adv-goodthe knife cuts sugar cane well

b. *ki-jiko ki-na-kul-a wali vi-zuri7-spoon 7SA-PRT-eat-FV 1 lrice Adv-goodthe spoon eats rice well

The "spoon" may not be the subject of this verb. Assuming VP-internal subjecthypothesis (Koopman and Sportiche, 1991), it can be seen that "knife" may begenerated in that subject position, but "spoon" may not. How does theapplicative construction handle such facts is a question that needs furtherresearch.

How are the surface features derived? For example, if all applied objectsare base-generated in the same spec position higher than the direct object,then we need to explain how the order of the objects is DO AP. We need toexplain also why there is a symmetry between the two objects inasymmetrical object languages such as Chichewa.

The typology, therefore, mainly is intended to help in organising the datawhich is to be studied. It is nevertheless a very important tool since it raisesinteresting questions regarding the grammar of diverse phenomena in theBantu languages that is related to applicatives. Such phenomena include verbextensions, constituent structure, agreement and quantification.

ReferencesAlsina, A. and S. Mchombo. 1990. The Syntax of Applicatives in Chichewa:

Problems for a Theta Theoretic Asymmetry. Natural Language andLinguistic Theory 8:493-506.

Ashton, E.O. 1966. Swahili Grammar. London: Longman.Baker, M. 1988a. Theta Theory and the Syntax of Applicatives in ChichevVa,

Natural Language and Linguistic Theory 6: 353-389.Baker, M. 1988b. Incorporation: A Theory of Grammatical Function Changing.

Chicago: Chicago University Press.Baker, M. 1990. Elements of a Typology of Applicatives in Bantu. In J.

Hutchison and V. Manfredi (eds) 1990. Current Approaches to AfricanLinguistics 7. Dordrecht: Foris.

3 1

Baker, M. 1992. Thematic Conditions on Syntactic Structures. In I. Roca (ed)Thematic Structure: Its Role in Grammar. Foris Publications: New York.

Barss, A. and H. Lasnik. 1986. A Note on Anaphora and Double Objects.Linguistic Inquiry 17: 347-354.

Bergvall, V. 1986. A Typology of Empty Categories in Kikuyu and Swahili.Dimmendaal, G.J. (Ed) Current Approaches to African Linguistics III.Dordrecht: Foris.

Bresnan, J. and L. Moshi. 1990. Object Asymmetries in Comparative Bantu.Linguistic Inquiry 21, 2:147-185.

Chomsky, N. 1986. Knowledge of Language. Praeger: New York.

Hale, K. and S. J . Keyser. 1993. On Argument Structure and the LexicalExpression of Syntactic Relations. In K. Hale and S. Keyser (eds) TheView from Building 20. Cambridge: MIT Press.

Hoffman, M. 1991. The Syntax of Argument Structure Changing Morphology.Ph.D. dissertation, MIT.

Hyman, L. and A. Duranti. 1982. On the Object Relation in Bantu, in P.Hopper and S. Thompson (eds). Studies in Transitivity (Syntax andSemantics 15). New York: Academic Press.

Kayne, R. 1994. The Antisymmetry of Syntax. Cambridge: MIT.Kimenyi, A. 1980. A Relational Grammar of Kinyarwanda. Berkeley:

University of California Press.Kinyalolo, K.K. 1991. Syntactic Dependencies and the Spec-Head Agreement

Hypothesis in KiLega. Ph. D. dissertation, UCLA.Koopman, H. and D. Sportiche. 1991. The Position of Subjects. Lingua 85:211-

258.Kural, M. 1993. Instrumentals and Causatives. ms. UCLA.Lakoff, G. 1968. Instrumental Adverbs and the Concept of Deep Structure.

Foundations of Language 4, 1:4:29.Larson, 1988. On Double Object Constructions. Linguistic Inquiry 19:335-91.Marantz, A. 1993. Implications of Asymmetries in Double Object

Constructions. In S. Mchombo (ed). Theoretical Aspects of BantuGrammar. Stanford: CSLI.

Mchombo, S. (ed) 1993. Theoretical Aspects of Bantu Grammar. Stanford:Center for thc Study of Language and Information.

Ngonyani, D. 1994. A Classification of Southern Tanzanian Languages. MAthesis, UCLA.

32

Otani, K. and J. Whitman. 1991. V-Raising and VP-Ellipsis. Linguistic Inquiry,22,2:345-358.

Trithart, M. L. 1983. The Applied Affix and Transitivity: A Historical Study inBantu. Unpublished Ph.D. dissertation, UCLA.

Deo NgonyaniLinguistics DepartmentUniversity of California Los Angeles405 Hilgard Ave.Los Angeles, CA 90095

ngonyani@humnet.ucla.edu

33 3 5